Codebases often disappoint us, but we keep creating new ones with the absurd hope that it will be different this time. Spoiler alert: it never is. Notably, research codebases need to fulfil the oxymoron of being robust for long and large computations and highly flexible to experiment with outlandish ideas quickly. But what if there was a way to break the cycle and build code that endures, even for groundbreaking research. This talk isn't about magic bullets or silver linings. It's about the powerful connection between healthy teams and healthy code. We'll explore how a healthy team improved the health of a codebase by introducing type annotations and runtime type checking for JAX arrays.